[R] errors in package nortest
Robert Offinger
Robert.Offinger at uni-passau.de
Mon Mar 29 12:11:02 CEST 2010
There are two errors in the R package nortest.
I tried to contact the maintainer Juergen Gross
but to no avail.
How should one progress if the mail address is still
valid but the maintainer does not answer or react in
any way, e.g. since he left the university? The
automatic R checking system for orphaned packages
will not catch such a situation, I guess.
And now on to the errors:
The first error in the function cvm.test was in a way
found by
http://tolstoy.newcastle.edu.au/R/help/06/06/29029.html
but the data set shows only the effect.
Here is the problem:
The approximation formula used in cvm.test has the
effect that the p-value increases for very large values
of the test statistic WW (WW>=1.334).
Therefore very large values of WW result in absurd
p values (as one can see in the given link).
The easiest approach would be to set the resulting
p-value to 0 for large values of WW (e.g. WW>=1.334)
The second error in the function lillie.test was described
in
https://stat.ethz.ch/pipermail/r-devel/2007-July/046282.html
but this error is really harmless since in
p <- pnorm((x - mean(x))/sd(x))
the case sd(x)==0 was not caught.
Robert Offinger
More information about the R-help
mailing list